Continuous EEG Signal Decoding with CCA Denoising

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current Human-Machine Interfaces (HMIs) rely on motor actions, introducing time lag due to nerve transmission delays, and existing Brain-Machine Interfaces (BMIs) based on EEG signals provide discrete outputs, which are not suitable for continuous tasks like vehicle steering and may not anticipate motor actions or account for confounding non-brain components.

Innovation Solution

A computer-implemented method using Canonical Correlation Analysis to separate confounding components from biopotential signals, followed by Multiple Linear Regression to decode continuous signals, allowing for short-term prediction and direct brain input, enabling continuous motor action commands or other signals like attention levels.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If classification-based decoding of motor actions from EEG signals is used, then discrete outputs are obtained, but continuous outputs required for tasks like vehicle steering cannot be achieved

Engineering Contradiction:
Improveoutput continuityVSAvoidtask applicability
Core Design Contradiction:
Measurement precisionVSEase of operation

Solution Approach 1:

The patent transforms the output parameter from discrete classification labels to continuous signal values by using Canonical Correlation Analysis to extract continuous neural signals that correlate with motor actions, enabling smooth control for tasks like vehicle steering

Inventive Principle:
Principle #35Parameter changes

2Reliability

If decoders are fitted from data recorded when users imagined performing motor tasks, then decoding models can be obtained, but the ability to anticipate motor actions is lost and confounding non-brain components remain in EEG signals

Engineering Contradiction:
Improvesignal accuracyVSAvoidanticipation capability
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The patent extracts and removes confounding non-brain components from EEG signals using Canonical Correlation Analysis, isolating the pure neural signals that carry motor action information while eliminating artifacts and unrelated physiological signals

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

By analyzing denoised neural signals with anticipated time shifts, the system performs preliminary detection of motor intentions before actual motor actions occur, enabling prediction and preparation for upcoming movements

Inventive Principle:
Principle #10Preliminary action

3Speed

If motor actions are used for inputting human commands, then direct control is achieved, but nerve transmission delay introduces time lag

Engineering Contradiction:
Improvecontrol response speedVSAvoidreaction time lag
Core Design Contradiction:
SpeedVSLoss of time

Solution Approach 1:

The system detects and decodes neural signals before motor actions are executed, allowing the control system to prepare and respond to user intentions in advance, thereby eliminating the time lag associated with nerve transmission and motor execution

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3912547A1Computer-implemented method and data-processing device for obtaining continuous signals from biopotential signals
Publication Date: 2021.11.24 TOYOTA JIDOSHA KK
  • EP3912547A1 patent drawingFigure 1~2
  • EP3912547A1 patent drawingFigure 3~4
  • EP3912547A1 patent drawingFigure 5

AI summary

The present invention concerns a computer-implemented method for obtaining continuous signals from biopotential signals. This method comprises the separation of confounding non-brain components from the biopotential signals by a statistical correlation analysis algorithm, such as a Canonical Correlation Analysis algorithm, to obtain denoised neural signals, and the decoding of the continuous signals from the denoised neural signals. The present invention also relates to a data-processing device (1) configured to execute this method.